Transcript NCPP – needs, process components, structure of scientific climate

NCPP – needs, process
components, structure of
scientific climate impacts study
approach, etc.
NCPP - Two main elements
identified so far
NCPP
Providing future downscaled
and value-added climate
information
Digital data
access
Translational
information
Documentation
Archiving and providing a
repository of best practices
standards/guidance, tools, etc.
Types of needs*, identified so far
•
# 1 - Supplying already existing downscaled data that fulfill the open-source
and review (quality) criteria in GIS format
• For non-standard areas – watersheds, states, other
•
•
# 2 – Re-gridding or interpolation to locations of already existing
downscaled data – tools and translational information
# 3 - Supplying sets of indices (value-added information) by sector as
needed for climate change impacts studies
• For different spatial scales – states, regions, cities, watersheds, ecological regions,
other
•
# 4 - Downscaling variables not available in existing downscaled data set
portals
– Standard available variables from existing portals – tmax, tmin, precipitation, on monthly or daily
scale
• Supplying this data in GIS format
• Re-gridding as necessary
• Supplying application-specific indices
•
•
# 5 – Supplying narratives relative to specific variables, and sectors and at
specific spatial scales – historical and for future periods (what has
happened, what is going to happen)
# 6 – Supplying translational information relative to downscaling
methodology, uncertainty of results, interpretation of results, re-gridding
procedures, etc.
* the needs are not listed based on importance or priority
• This presentation focuses first on:
– Details related to need # 3 - Supplying sets of
indices (value-added information) by industry
as needed for climate change impacts studies
– Details related to need # 4 - Downscaling
variables not available in existing downscaled
data set portals
Scientific approach to climate
change impacts study
• Some definitions
– Index or measure - a number derived from
observations or simulations
– Growing Degree Days, Heating and Cooling Degree Days,
frequency and intensity of heat waves, Cold spells, date of first
fall frost and last spring frost, exceedance probability of annual
precipitation, persistence of rain/dry days, etc.
– Ensemble - A group of parallel model simulations
used for climate projections.
– Variation of the results across the ensemble members gives an
estimate of uncertainty.
– Ensembles made with the same model but different initial
conditions only characterize the uncertainty associated with
internal climate variability
– Multi-model ensembles including simulations by several models
also include the impact of model differences.
Starting stage – need # 3
Structure of a climate impacts study - focus on process
when using readily available downscaled data
Example: Historical and projected future precipitation variability
in the Colorado River Basin
Comparison of the model simulations and
the observational data sets
for the historical period 1951-1999
Evaluation of the precipitation variability
during 2001-2099
Assessment of the quality of the
observed gridded data sets
for the 1951-1999 period
Structure of a climate impacts study - focus on process when
using readily available downscaled data
Example: Historical and projected future precipitation variability
in the Colorado River Basin
Assessment of the homogeneity of the
observed gridded data sets
for the 1951-1999 period
Comparison of the model simulations and
the observational data sets
for the historical period 1951-1999
Evaluation of the precipitation variability
during 2001-2099;
Inclusion of translational information –
how to interpret the results and the uncertainty
Quality assessment of
observed data;
Inclusion of translational
information
Bias evaluation of
the GCM downscaled data;
Inclusion of translational
information – how to interpret
the biases
Results may impact:
Set of models used in future
precipitation changes analysis
(choice of best models)
Interpretation of results
(choice of interpretation based
only on non-biased models)
Evaluation of the precipitation variability
during 2001-2099
Description of
measures’ calculation
Calculated set of measures of precipitation variability
Mid future period
2026-2074
Early future period
2001-2049
Downscaled
by L. Brekke
et al.
A2
B1 A1b
Downscaled
by J. Eischeid
(NOAA)
A1b
Downscaled
by L. Brekke
et al.
A2
B1 A1b
Downscaled
by J. Eischeid
(NOAA)
A1b
Late future period
2051-2099
Downscaled
by L. Brekke
et al.
A2
B1 A1b
Downscaled
by J. Eischeid
(NOAA)
A1b
Calculate deltas (changes) for all GCM future periods versus the GCM historical period
Qualitative or quantitative comparison of the future changes between periods, SRES emissions scenarios,
and downscaling methodologies
Description of deltas calculations, comparison methodology, interpretation of results and uncertainty
Summary - Using already downscaled data sets –
for ex., Maurer et al., Hayhoe, NARCCAP, other
Access to downscaled GCM and/or RCM data and to observed data:
Data set 1 – multiple
Data set N – multiple
GCMs, SRES emissions
scenarios
GCMs, SRES emissions
scenarios
Observed data set 1, 2
Quality control, homogeneity,
testing, if needed
Calculation of measures/indices
Validation - comparison between GCM and observed
measures/indices during historical period of overlap
Feedback
from users
Translational
info about
GCM biases
Products: a) future changes in various measures/indices
of interest for time segments or overlapping periods, with uncertainty
assessment, b) narratives, c) probabilistic distributions of
measures/indices, d) other
Translational information: Interpretation guidance of results,
Uncertainty around the measures/indices, Details of
Probability distributions’ methodology
Examples of existing data sets for
the US that can be used:
•
Observed data sets
– Gridded data:
• Maurer et al. 2002 – daily and
monthly; ~12 km resolution
• PRISM – Precipitation Regressions
on Independent Slopes Model, (Daly
et al 2004, 2006) –monthly, 4 km or
less resolution
– Station data
• Located at NCDC or at state
climatologist offices:
– Data from COOP stations - daily,
monthly
– Data from the Historical Climatology
Network (homogenized) – monthly
•
Downscaled data sets:
– Gridded data:
• Maurer et al. 2007 CMIP3 BCSD
downscaled monthly data set,
approx. 12km, 1950-2099, 16
GCMs, 36 projections, 3 SRES
scenarios
• Hayhoe downscaled daily data set
– to be serviced by USGS –
approx. 12km, tmax, tmin, precip,
1960-2099, 4GCMs, 4 SRES
scenarios, CONUS and Alaska
• NARCCAP dynamically
downscaled data set
• CMIP3 BCCA downscaled daily
data set – same spatial domain as
BCSD; same resolution; 3 time
periods 1961-2000, 2045-2064,
and 2080-2099; tmax, tmin, precip
• Project portals at institutions
– Station or location data
• At Universities, institutions
Generalization - Using already downscaled data
sets – for ex., Maurer et al., Hayhoe, NARCCAP, other
Components of the process
Access to downscaled GCM and/or RCM data
Access to observed data:
Quality control tool
Calculation of measures/indices
Validation of measures/indices tool
Feedback
from users
Ensemble analysis tool; Uncertainty analysis tool;
Probabilistic distributions tool, other
Products
Translational information tool
Translational
Information
tool
Later stages
Structure of a climate impacts study - focus on
process when downscaling GCM and/or RCM data
The main differences are related to the beginning of the process
and the access to data portals of raw GCM and/or RCM data and
the subsequent downscaling of the GCM or RCM data.
The subsequent slides focus on the downscaling procedures when using different
downscaling techniques.
S
T
A
R
T
L
A
T
E
R
Access to downscaled GCM and/or RCM data and to observed data:
Data set 1 – multiple
Data set N – multiple
GCMs, SRES emissions
scenarios
GCMs, SRES emissions
scenarios
Observed data set 1, 2
Downscaling GCM and/or RCM data; Access to observed data:
Data set 1 – multiple
Data set N – multiple
GCMs, SRES emissions
scenarios
GCMs, SRES emissions
scenarios
Observed data set 1, 2
Structure of a climate impacts study – focus on downscaling
process (statistical or empirical –dynamical downscaling)
User identified requirements for the downscaled product –
Measures/indices of interest, temporal and spatial scales
User and provider agreed upon sources of uncertainty
Set of GCMs, SRES emissions scenarios, downscaling techniques
Access to observed data sets
of the variables of interest and predictors
(if needed) - quality controlled, homogeneous
Develop transfer functions for a given period
Validate transfer functions on separate period
Comparison indicates downscaling biases
Access to raw GCM data sets
of the variables of interest and
of the predictors (if needed)
Apply transfer functions on
GCM predictor data for
control period
Validate CGM transfer functions
for control period vs OBS data
Comparison indicates GCM + downscaling biases
Apply transfer functions on GCM future data to obtain downscaled projections
Products - Calculate additional indices of interest, potential changes, or
downscaled data needed for input in process models, other
Structure of a climate impacts study – focus on downscaling
process (dynamical downscaling)
User identified requirements for the downscaled product –
Measures of interest, temporal and spatial scales
User and provider agreed upon sources of uncertainty
Set of GCMs, SRES emissions scenarios, downscaling techniques
Definition of
“observed”,
“current”,
“control” and
“future” climates
for RCM
simulations and
the types of
comparisons
that must be
performed. Note
that it is not
appropriate to
compare future
climate
projections
directly to
observations.
(Winkler et al.,
2011)
Products:
time series,
climatologies,
potential changes
for future time slices,
other
Structure of a climate impacts study – focus on downscaling
process - Example – Maurer et al. 2007 downscaling effort
(disaggregation downscaling)
Observed gridded
data
Step 1
Bias
correction
GCM 20th century
data
GCM 20th century
data
Re-grid observed and GCM 20th and 21st century data
to same resolution (2º)
For each grid cell, create monthly cumulative distribution functions
(CDFs) by variable of interest
Create quantile map
By grid cell and for each month, adjust the 20th and 21st century GCM data
by using the OBS values for given quantiles of the CDF
Step 2
Spatial
Downscaling
Compute
Factor values
by grid cell
and time step
Interpolate
2º factor
values to 1/8º
Apply factor values to
original 1/8º observed data Resulting in BSCD GCM data
END
Comparison of the model simulations and
the observational data sets
for the historical period 1951-1999
Observed data sets
GCM data sets
Calculated variables/measures compared to:
Maurer et al. 2002
Gridded observed data
Downscaled by
L. Brekke et al.
(2007)
37
model
runs
PRISM (Daly et al.
2004, 2006) Gridded
Observed data
Downscaled by
Jon Eischeid
(NOAA)
30
model
runs
Translational
info about the
data sets;
the comparison
methodology;
about the
interpretation of
the biases of the
downscaled
data;